摘要
针对传统感应信号控制的初始绿灯时间和单位绿灯延长时间设置单一导致行车延误增加的问题,在分析信号交叉口绿灯放行时间与绿灯需求时间差异因素的基础上,根据交叉口车辆到达特征和车型,构建了基于LSTM神经网络的信号交叉口排队车辆车头时距预测模型,对绿灯放行时间进行精确计算,使之能灵活适应实际车流特征.分析结果表明:LSTM神经网络预测模型的平均绝对百分比误差(MAPE)为0.17,均方根误差(RMSE)为0.16,预测值与真实值拟合度好.在案例中,相比较于传统感应信号控制的交叉口,采用该模型的改进效果显著,所提高的通行效率可达14.89%.
The initial green time and the unit extension time of traditional actuated signal control system was found lack of variability,which resulted in traffic delays.Based on the analysis of the mismatch between green light release time and green light demand time,according to the vehicle arrival condition and traffic composition,a headway prediction model based on the long short-term memory(LSTM)neural network was constructed,the proposed model could effectively predict headway of queuing vehicles.The initial green time and the green light extension time of each phase were calculated to make signal flexibly adapt to the actual traffic flow.The results showed that the mean absolute percentage error(MAPE)of the prediction model of headway was 0.17,and the root mean square error(RMSE)was 0.16.The predicted value was basically consistent with the true value.Compared with the traditional actuated signal control,the improved actuated signal control system increased the performance efficiency by 14.89%.
作者
毛程远
王琴
钱俊
姜伟
求英浩
杨胜德
张欣环
MAO Chengyuan;WANG Qin;QIAN Jun;JIANG Wei;QIU Yinghao;YANG Shengde;ZHANG Xinhuan(College of Engineering,Zhejiang Normal University,Jinhua 321004,China;Modern Logistics Development Administration Center of Jinhua,Jinhua 321000,China;Shengzhou Natural Resources and Planning Bureau,Shengzhou 312400,China)
出处
《浙江师范大学学报（自然科学版）》
CAS
2022年第2期210-217,共8页
Journal of Zhejiang Normal University:Natural Sciences
基金
浙江省自然科学基金资助项目(LY18G030021,LY21G010005,LY18G010009)
浙江省教育厅科研项目(Y201738488)。
关键词
智能交通
感应信号
短周期预测
车头时距
神经网络
intelligent transportation
actuated signal control
short period prediction
time headway
neural network
